I honestly don't understand how to make at least someone find out about my AI service.

Radja Echpochmakov — Wed, 09 Sep 2026 13:04:17 +0000

I've been doing and continuing to do an AI-related project for a month and a half (and yes, I've also seen all these posts saying that the era of AI is already passing away, etc., but as I assume, this is not the case, this kind of "intelligence" is simply evolving financially and marketing terms to become an integral part of life for a person, like the same calculator... okay, I digress).

The project in question:

just-not-google / BiNeuron

BiNeuron is a local AI assistant that detects programming languages from requests/files, selects specialized models based on hardware, and generates code/docs/analysis via configurable prompts. It extracts text from PDFs, Word, images (OCR), and other formats, translates, filters profanity, and uses proxies if needed. Runs offline.

BiNeuron

🇬🇧 English

Intelligent Code Analysis and Generation Platform

BiNeuron is a sophisticated software solution that bridges the gap between human intent and machine‑generated code. It unifies advanced natural language processing, optical character recognition, and adaptive model selection into a single, powerful tool designed for developers, researchers, and technical teams.

At its core, BiNeuron automatically identifies the programming language of a given request, extracts content from a wide array of file formats—including images and documents—and then generates context‑aware, production‑ready code using best‑in‑class local or cloud‑based language models.

Key Capabilities

Programming Language Detection

Supports over 25 programming languages, including Python, Java, C/C++, C#, JavaScript, TypeScript, Go, Rust, Swift, Kotlin, Ruby, Dart, Julia, Lua, SQL, MATLAB, R, Pascal, Assembly, Fortran, F#, Ada, Zig, PHP, Shell, Scala, and more.
Combines heuristic algorithms, proprietary keyword matching, and AI‑powered orchestration to achieve high detection accuracy.
Analyzes both user‑supplied text and the content of attached files…

View on GitHub

The idea of my project is to run AI for coding (vibcoding) completely locally, because after reading the news in the IT world (I don't read other news), I realized that there would be no more ordinary writing of code as it is now. For example, there are already many tasks of a routine and similar nature of the transition to "development" to AI and neural networks (why did I put development in quotation marks? It's simple, due to the fact that so far many AI models hallucinate and add context to the main task, and as a result, the output is something between "thank you, I saved 40 minutes of my life" and "what the hell is this, I've been trying to explain for 20 minutes that I need the usual list and nothing more). And so I decided to enter this niche, but first of all not for the sake of money (well, as for me, it doesn't smell like it for the next 2 years at all), but purely because I see that this market is not really developed either in the West or in our country. For example, what services can you call "vibcoding", originating from Russian IT offices and without using ready-made APIs from foreign LLMs. For example, I know only one such project, VibeCraft, and it's from Yandex, the monopolist of the Runet, so it doesn't count.

If we continue talking about my project, its main features are:
1) Local launch, which was mentioned above;
2) The complete absence of cloud APIs of any models;
3) Built-in crawling methods for downloading the model, such as a site mirror and everyone's favorite proxy (how much pain there has been in this word for the last 2 years.. Ehh)
4) Automatic mode, in which the application itself determines how powerful your PC or laptop is through algorithmic calculations and measurements. As well as defining and downloading the necessary AI model that is highly specialized in one specific programming language (or multilingual, if several programming languages are required to achieve the task)
5) A large catalog of file support, as a context that you can add to AI to solve your problem (including reading text from pictures)
6) Changing files using AI, but this feature is disabled by default so that there are no cases like: "Claude Code erased 700 GB of data, but it turned out to be important for the future life of the project"
7) Virtual storage is the directory selected as the main one on your computer, so that each time you do not attach files to the AI, the application automatically takes all the files from this directory and reads them.
8) And much more, various features and little things that make the overall use of the app much better

So I didn't know that the easiest part of this whole project would be its technical part (writing code). Even though before this project, I was like a "cow on ice" in Machine Learning and related sciences, that is, I was clumsily trying to do something.

But the most difficult thing, without any confusion, I can say - it was and is PR and marketing. Because I sincerely do not understand how recommendation algorithms work and whether there are any recommendations at all on the part of those who publish, not just those who read, or whether these are just useless "dancing with a tambourine". And besides, if I had at least some budget for promotion, it would have been a little easier. Many people are skeptical, not just of direct articles like "I made the coolest AI app", but even of simple technical articles on project promotion and development, if at the end of a long and interesting post the author put a link and asked just to see what he described with all this text above.

One of the interesting things about this project is its idea and the main feature with AI-orchestrated models, it was invented as a result of ordinary thoughts before going to bed and nothing more, no beautiful story. Well, the only thing that can be noticed is that before this idea, I came up with and rejected more than 40 ideas.

Well, that's where my ADHD-like retelling of the product comes to an end, I became too lazy to continue this post, so there are too many letters already in it, and after all, I'm a buzzer and chronically can't digest more than 2 words... Good luck to everyone and bye!

P.S. If anyone is interested, this project is currently completely open in terms of code, as well as absolutely free.

DEV Community: Radja Echpochmakov

I honestly don't understand how to make at least someone find out about my AI service.

just-not-google / BiNeuron

BiNeuron

Key Capabilities

Programming Language Detection